can drmemory, and should drmemory, support traces?

_From [bruen...@google.com](https://code.google.com/u/109494838902877177630/) on March 06, 2012 16:42:31_

xref shift issue from data labels: issue #784 originally I was worried about data labels and decode_fragment() for traces: but that was in the old DR API. the new DR API passes the bbs to the bb event again even if there's no trace event callback.

here are some of my old notes:
*\* TODO should be able to turn on traces after all?
- see whether apps that traces help are helped w/ drmem in place
- see whether can do better scratch reg picking w/ fewer spills/restores
over whole trace

actually we can't turn on traces: we assume that dr_fragment_exists_at()
will find all code coming from a call to a retaddr.
we also assume post_call_table() contains only tags (issue #114)
and that we can store just the size and have contiguous fragments (issue #260).

update to those notes: the issue #114 concern is no longer there b/c issue #673 changed the invalidation to be on re-use instead of delete (though the exists-at is still there and is relevant to drwrap and not just drmem)

_Original issue: http://code.google.com/p/drmemory/issues/detail?id=818_